Interleukin-3 Binding to the Murine βIL-3 and Human βc Receptors Involves Functional Epitopes Formed by Domains 1 and 4 of Different Protein Chains*

Interleukin-3 (IL-3) is a cytokine produced by activated T-cells and mast cells that is active on a broad range of hematopoietic cells and in the nervous system and appears to be important in several chronic inflammatory diseases. In this study, alanine substitutions were used to investigate the role of residues of the human β-common (hβc) receptor and the murine IL-3-specific (βIL-3) receptor in IL-3 binding. We show that the domain 1 residues, Tyr15 and Phe79, of the hβc receptor are important for high affinity IL-3 binding and receptor activation as shown previously for the related cytokines, interleukin-5 and granulocyte-macrophage colony-stimulating factor, which also signal through this receptor subunit. From the x-ray structure of hβc, it is clear that the domain 1 residues cooperate with domain 4 residues to form a novel ligand-binding interface involving the two protein chains of the intertwined homodimer receptor. We demonstrate by ultracentrifugation that the βIL-3 receptor is also a homodimer. Its high sequence homology with hβc suggests that their structures are homologous, and we identified an analogous binding interface in βIL-3 for direct IL-3 binding to the high affinity binding site in hβc. Tyr21 (A–B loop), Phe85, and Asn87 (E–F loop) of domain 1; Ile320 of the interdomain loop; and Tyr348 (B′–C′ loop) and Tyr401 (F′–G′ loop) of domain 4 were shown to have critical individual roles and Arg84 and Tyr317 major secondary roles in direct murine IL-3 binding to the βIL-3receptor. Most surprising, none of the key residues for direct IL-3 binding were critical for high affinity binding in the presence of the murine IL-3 α receptor, indicating a fundamentally different mechanism of high affinity binding to that used by hβc.

Interleukin-3 (IL-3) 1 is a cytokine produced by activated T-cells and mast cells that has been shown to stimulate re-newal of pluripotent hematopoietic stem cells and to be a potent regulator of many hematopoietic cell lineages (1)(2)(3). Its role appears to be in stimulating inducible hematopoiesis in response to parasite infections (4), and it has also been implicated in the pathogenesis of several chronic inflammatory diseases, including asthma (1), and neurodegenerative disorders, such as multiple sclerosis (5). The effect of IL-3 on human cells is mediated by a receptor system composed of a ligand-specific ␣ subunit and a ␤ subunit (denoted ␤c) that is also part of the receptor systems for the related cytokines, interleukin-5 (IL-5) and granulocyte-macrophage colony-stimulating factor (GM-CSF) (6 -9). Signaling through the ␤c receptor requires the formation of a high affinity complex involving each cytokine and its respective ␣ subunit (7)(8)(9)(10). Whereas the ␣ subunits bind their ligands with low affinity, ␤c does not measurably bind any of the ligands alone. Upon receptor activation, the cytoplasmic portion of the ␤c subunit, which lacks any intrinsic kinase activity (11), initiates a number of signaling pathways including the Janus kinase 2/signal transducers and activators of transcription, phosphatidylinositol 3-kinase, and Ras/mitogen-activated protein kinase pathways (reviewed in Ref. 12).
Mice also possess a ␤c subunit (m␤c) but have an additional IL-3-specific ␤ receptor (␤ IL-3 ). ␤ IL-3 differs from the m␤c subunit in its ability to bind murine IL-3 (mIL-3) directly (13), although the presence of the mIL-3 ␣ subunit is absolutely required for signaling (14). The properties of mIL-3-responsive precursor cells from gene knock-out mice lacking expression of the ␤ IL-3 subunit indicate that this subunit plays an important role in the response to mIL-3 stimulation (15).
The human ␤c (h␤c), m␤c, and ␤ IL-3 receptors are closely homologous in sequence and belong to the hematopoietin (or class I cytokine) receptor superfamily. Members of this family have extracellular domains containing the cytokine-receptor homology module composed of two fibronectin III domains that have a number of conserved sequence elements (16,17). X-ray structures of ligand-receptor complexes of simpler members of the family, including growth hormone (18), erythropoietin (19) and interleukin-4 (20), have demonstrated that in each case the structural epitopes for ligand binding involve loops at approximately the orthogonal interfaces formed between the two fibronectin III domains of these receptors. In addition, mutagenesis studies of these receptors have shown that within the structural epitopes only a relatively small number of residues, which compose the functional epitope, are critical determinants of ligand binding (21)(22)(23)(24)(25). These residues are contributed by a combination of the A-B and E-F loops of the membrane-distal fibronectin III domain and the BЈ-CЈ and FЈ-GЈ loops of the membrane-proximal domain.
To what extent the principles of ligand binding established for the two domain receptors apply to the h␤c receptor and its close relatives has been unclear because the h␤c receptor does not engage ligand in the absence of the ␣ subunits and possesses an extracellular domain with two cytokine-receptor homology modules. Residues in the domain 4 BЈ-CЈ and FЈ-GЈ loops have been identified previously (28 -30) by mutagenesis studies modeled on the growth hormone receptor as critical for formation of a high affinity ligand binding complex. Prior to the structure of the h␤c receptor being elucidated, it was expected that domain 3 would partner with domain 4 in forming the ligand-binding interface, but no binding residues in domain 3 of the h␤c receptor were identified. The h␤c receptor structure gave new insight into the ligand-binding interface thus predicting the involvement of domain 1. The crystal structure of the complete extracellular domain of the h␤c subunit showed that h␤c exists as a preformed homodimer (Fig. 1), in which the G-strands of domains 1 and 3 are swapped between the two monomers (26). As a result of the intertwining of two monomers, domains 1 and 4 of symmetry-related chains form an approximately orthogonal interface that resembles the ligandbinding interfaces of two-domain hematopoietin receptors. The participation of domain 1 in ligand binding has been supported by recent mutagenesis studies that have shown that residues in the A-B and E-F loops of domain 1 of h␤c are required for the formation of the high affinity human GM-CSF (hGM-CSF) complex and for receptor activation by hGM-CSF and human IL-5 (hIL-5) (31). The involvement of domain 1 also provides further evidence for the biological relevance of the h␤c receptor structure. The murine ␤ IL-3 subunit has the interesting feature of binding mIL-3 directly with low affinity as well as forming a high affinity complex in the presence of the mIL-3␣ receptor (14). It has been shown that the carboxyl-terminal cytokine-receptor homology module of ␤ IL-3 makes a unique contribution to the direct binding of mIL-3 and that the domain 4 BЈ-CЈ loop has a critical role (32). The amino-terminal cytokine-receptor homology module is also essential, but the corresponding portion of m␤c can effectively substitute for that of ␤ IL-3 without compromising direct mIL-3 binding (32). However, a detailed analysis of the residues involved in direct and high affinity IL-3 binding has not been carried out previously nor has the structure of the ␤ IL-3 subunit extracellular domain been determined.
In the present work, we show by using analytical ultracentrifugation and cross-linking that the expressed extracellular domains of the m␤c and ␤ IL-3 receptors, like h␤c, are homodimers. This, together with the high sequence homology between the murine receptors and h␤c, suggests that both m␤c and ␤ IL-3 are likely to have intertwined homodimer structures analogous to that determined for h␤c (26). Consistent with this prediction, we demonstrate here that domains 1 and 4 cooperate to form the functional epitope for direct IL-3 binding to the murine ␤ IL-3 receptor and similarly form the high affinity IL-3-binding site of the h␤c receptor. Most surprising, however, we have shown that the residues in the "elbow" region of the murine ␤ IL-3 receptor have little effect on the high affinity binding of mIL-3, indicating that this receptor forms a high affinity complex in a different way to the h␤c receptor.

MATERIALS AND METHODS
Isolation of Murine ␤c, ␤ IL-3 , and IL-3␣ cDNAs-cDNAs encoding the extracellular domains of the m␤c subunit, the extracellular domain, the full-length ␤ IL-3 subunit, and the full-length mIL-3 ␣ subunit were isolated from FDC-P1 total RNA by reverse transcription using Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA). The sequences of these cDNAs were verified using Big Dye terminator cycle sequencing (Applied Biosystems, Gladesville, Australia).
Site-directed Mutagenesis of ␤ IL-3 cDNA-Site-directed mutagenesis was performed by using the QuikChange method (Stratagene) using Pfu Turbo DNA polymerase. The complete sequences of mutant ␤ IL-3 subunits were verified by Big Dye Terminator cycle sequencing (Applied Biosystems).
Expression and Purification of h␤c, m␤c, and ␤ IL-3 Extracellular Domains-cDNAs encoding the extracellular domain of the h␤c, m␤c, and ␤ IL-3 subunits were subcloned into the EcoRI site of pBacPAK8 for baculovirus expression. Recombinant baculovirus was produced, and protein was expressed and purified essentially as described previously (34). Briefly, concentrated baculovirus supernatant, containing expressed protein, was applied to a Sephacryl S-200 (Amersham Biosciences) gel filtration column, and fractions containing protein were concentrated before being chromatographed on a MonoQ HR 5/5 column (Amersham Biosciences) at pH 6.5 using a fast protein liquid chromatography apparatus and eluted using a 0 -350 mM NaCl gradient. Protein concentrations were determined using a Bradford assay (Bio-Rad), with bovine ␥-globulin as the protein standard. The identities of these proteins were confirmed by amino-terminal sequencing, which was performed as described previously (34).
Chemical Cross-linking of ␤ Subunit Extracellular Domains-Purified h␤c, m␤c, and ␤ IL-3 extracellular domain proteins (1 g), in 0.1 M sodium phosphate buffer (pH 7.5), were reacted with 2 mM BS 3 (Pierce) at 25°C for 1 h, before the reaction was quenched with glycine (50 mM). Nonreducing SDS-PAGE was carried out by using 8% Tris-glycine gels with silver staining, as described previously (34).
Sedimentation Equilibrium Analysis of ␤ Subunit Extracellular Domains-Sedimentation equilibrium experiments were performed in a Beckman Optima XL-A analytical ultracentrifuge at 10,000 or 20,000 rpm by using 12-mm path length cells with carbon-filled double-sector centerpieces. The solution sector contained 100 l of protein sample at a concentration of ϳ1 mg/ml dissolved in 20 mM bis-Tris propane, pH 6.5, with 110 l of the solvent sector in the same buffer as reference. All experiments were performed at 25°C for 20 h. Scans at sedimentation equilibrium of absorbance (A) versus radial distance (r) in centimeters from the axis of rotation were collected at 280 nm. Scans at 360 nm were collected and subtracted from the equilibrium scans to correct for anomalies arising from cell windows. The resulting A versus r data was analyzed by using software supplied with the instrument. Where the data indicated that the sample was homogeneous, the molecular weight was evaluated by employing the sedimentation equilibrium shown in Equation 1, where R is the gas constant; T is the absolute temperature; is the angular rotation (radians/s), v is the partial specific volume of the protein (0.725 ml/g in this case), and is the density of the buffer. The molecular weights of h␤c and v were determined by running one sample of the protein in H 2 O and one in D 2 O in the sample experiment and solving the resulting two simultaneous equations. The best fit molecular weight for each sample was obtained using the program XLAEQ, and the homogeneity of the molecular weight was checked by evaluating point average molecular weights using the program XLAMW. The percentages of dimer and monomer in the samples were calculated using the program MULTMX1B.
Cytokines and Radiolabeling-mIL-2 and mIL-3 were produced by using the baculovirus expression system (34). Recombinant hIL-3 was purchased from Sigma. Purified mIL-3 or hIL-3 was radiolabeled with 125 I using the method of Fraker and Speck (35). Briefly, 6 g of mIL-3 or hIL-3 in 30 l of 0.1 M sodium phosphate buffer, pH 7.5, and 10 l (1 mCi) of Na 125 I in NaOH solution (Amersham Biosciences) were added to a microcentrifuge tube plated with 1 g of IODO-GEN reagent (Pierce) and reacted on ice for 10 min. Radiolabeled cytokine was purified on a Sephadex G-25 column (Amersham Biosciences) and stored at 4°C for up to 9 days.

Equilibrium Binding Analysis for hIL-3 and mIL-3-Hot saturation
binding experiments were performed on COS7 cells transfected with DNA constructs encoding the relevant receptors, as described previously (31). In order to assess the capacity of ␤ IL-3 subunits to bind mIL-3 directly, cold competition assays were performed on COS7 cells transfected with cDNAs encoding wild-type or mutant ␤ IL-3 subunits in the absence of the mIL-3␣ cDNA. 1 nM 125 I-labeled mIL-3 was added to 10 6 cells in 200 l of binding medium (RPMI 1640 supplemented with 0.5% w/v bovine serum albumin and 10 mM HEPES) with a serial dilution from a 200-fold excess of unlabeled mIL-3. Nonspecific binding was determined by performing the assay in the presence of 1 M unlabeled mIL-3. After incubation for 2.5-3 h at 4°C with intermittent agitation, the assay was terminated by centrifugation through 2:1 v/v dibutyl phthalate/dinonyl phthalate at 12,000 ϫ g for 4 min. The tips of tubes and visible cell pellets were counted using a Packard 5780 Auto-gamma counter. The dissociation constants (K d ) were determined from specific binding data by using the programs EBDA (36) and LIGAND (37) as described (31).
Proliferation Assays-[ 3 H]Thymidine incorporation into the DNA of CTLL-2 cells expressing wild-type or mutant hIL-3 or mIL-3 receptors was used to measure the capacity of these receptors to deliver a proliferative signal in response to hIL-3 or mIL-3, respectively, as described (34).

RESULTS
The ␤ IL-3 and m␤c Receptor Extracellular Domains Are Homodimers-Our studies of the h␤c receptor extracellular domain expressed in the baculovirus system have demonstrated that it exists as a preformed homodimer both in solution and when crystallized (26). The full-length receptor has also been verified to exist as a dimer in the cell membrane (26). The novel intertwined dimer structure of the h␤c receptor and the important ligand-binding loops of the elbow region formed by domains 1 and 4 of the two different protein chains are shown in Fig. 1. An alignment between the ␤ IL-3 and m␤c receptors and the h␤c receptor is shown in Fig. 2. The figure demonstrates the high degree of sequence homology between the extracellular domains of the murine and human ␤ receptors and the conservation of the cysteine residues known to form intramolecular disulfide bonds in h␤c. It is therefore likely that they have homologous structures.
To investigate this further, we carried out cross-linking and analytical ultracentrifugation studies of the purified extracel-lular domains of the ␤ IL-3 and h␤c subunits expressed in the baculovirus system. The m␤c subunit was also included for comparison. The amino-terminal sequences of the mature forms of ␤ IL-3 and m␤c had not been determined previously and were shown to be HEVTEEEETV and HGVTEAEETV, respectively (as shown in Fig. 2). The amino-terminal sequence of h␤c has been determined previously (34). Chemical cross-linking of all three ␤ subunit extracellular domains using the watersoluble reagent, BS 3 , gave analogous cross-linked products ( Fig. 3) supporting homologous structures.
Analytical ultracentrifugation studies were also carried out (Table I). Sedimentation equilibrium experiments were performed at pH 7.4 on the purified extracellular domains of ␤ IL-3 , m␤c, and h␤c. The average molecular weights determined for both m␤c and ␤ IL-3 were 83,000, compared with the average molecular weight of 91,300 obtained for h␤c from HL60 eosinophils and 87,500 for h␤c from TF1 cells. Because the predicted molecular weights (excluding glycosylation) of the ␤ IL-3 , m␤c, and h␤c(HL60) extracellular domains are 48,181, 48,331, and 47,739, respectively, the ultracentrifugation data indicate that the proteins exist as dimers at pH 7.4. There was some evidence of lower molecular weight protein in the samples. Although there was no evidence that this material was the monomer form of the subunits, fitting the data by using the program MULTMX1B showed that the monomer, if present, would constitute less than 10% of the concentration of the dimer.
Ultracentifugation experiments were also performed on the h␤c extracellular domain at pH 3 ( Table I). The average molecular weight obtained was 42,400 with some evidence of higher molecular weight material. These data indicate that the h␤c extracellular domain dissociates to the monomer at pH 3. The calculated frictional ratio for the monomer was 3.9, suggesting that denaturation had occurred at this pH in addition to dissociation of the dimer. The corresponding value for the dimer was 1.32.
Thus, the cross-linking and analytical ultracentrifugation studies indicate that the ␤ IL-3 and m␤c receptors are homodimers like h␤c. Based on the structure determined for h␤c, together with the alignment in Fig. 2, we were able to make detailed predictions of the likely secondary structure of ␤ IL-3 and the residues in potential ligand-binding loops.
Residues of the ␤ IL-3 Receptor Required for Direct IL-3 Binding-The functional epitope for ligand binding and activation of the h␤c receptor by hGM-CSF and hIL-5 includes Tyr 15  The loops implicated in human GM-CSF and IL-5 binding and receptor activation are labeled. Sticks depict Nlinked carbohydrate and are colored by atom type (green, carbon; blue, nitrogen; red, oxygen). This figure was drawn using PyMOL (27), and coordinates are available from the Protein Data Bank (accession code 1gh7). and the domain 4 residues Tyr 347 , His 349 , Ile 350 (BЈ-CЈ loop) (28,29), and Tyr 403 (FЈ-GЈ loop) (30). By analogy with human ␤c, we prepared alanine substitution mutants of residues in the relevant loops of the ␤ IL-3 subunit, and we examined their capacity to bind mIL-3 directly. Alanine substitution mutagenesis has been used extensively in the determination of ligandbinding sites in hematopoietin receptors, because alanine is found commonly in solvent-exposed and buried positions and in all types of secondary structure, and the methyl side chain does not cause steric or electronic interference to the protein chain. COS7 cells were transfected with expression vectors encoding wild-type or mutant ␤ IL-3 subunits and the resulting transiently transfected cells used in cold competition binding as-says as described under "Materials and Methods." Wild-type ␤ IL-3 was found to bind mIL-3 with a dissociation constant of 18.4 nM (Table II and Fig. 4). This is consistent with the K d values determined previously (13) by using a similar cold competition binding assay.
Involvement of Loops in Domain 1 of the ␤ IL-3 Subunit in Low Affinity mIL-3 Binding-The predicted A-B loop consists of residues Tyr 21 , Thr 22 , Asn 23 , and Arg 24 . Each of these residues was individually mutated to alanine, and the ability of these mutants to bind directly mIL-3 was assessed. The Y21A mutant ␤ IL-3 receptor exhibited no detectable binding, whereas the other three mutants were capable of mIL-3 binding with K d values within 2-fold of the K d for wild-type ␤ IL-3 (Table II). These results indicate that Tyr 21 is a critical binding determinant and the only significant residue in the A-B loop involved in mIL-3 low affinity binding.
The predicted E-F loop consists of the residues Tyr 82 , Thr 83 , Arg 84 , Phe 85 , Ser 86 , and Asn 87 . Mutant ␤ IL-3 subunits were prepared containing individual alanine substitutions of these residues, and their ability to bind mIL-3 with low affinity was  assessed (Table II). Both Y82A and S86A mutant ␤ IL-3 subunits exhibited impaired mIL-3 binding with an ϳ4-fold reduction in mIL-3 binding compared with wild-type ␤ IL-3 . The R84A mutant ␤ IL-3 was more severely affected, with ϳ13-fold less affinity than wild type. The F85A and N87A mutant ␤ IL-3 subunits did not detectably bind mIL-3. These results clearly implicate Phe 85 and Asn 87 as key direct mIL-3 binding determinants with significant secondary involvement of Arg 84 .
Involvement of the Interdomain and Domain 4 Loops of the ␤ IL-3 Subunit in Low Affinity mIL-3 Binding-Alanine-scanning mutagenesis was performed on the interdomain loop predicted to connect domains 1 and 4. This loop is made up of the residues Tyr 317 , Tyr 318 , His 319 , and Ile 320 . Mutant ␤ IL-3 subunits were prepared in which each residue was substituted with alanine, and the effect of these mutations on mIL-3 binding was assessed (Table II). The H319A ␤ IL-3 receptor was capable of wild-type low affinity mIL-3 binding, whereas the Y318A mutant exhibited ϳ3-fold reduction in mIL-3 binding. Low affinity mIL-3 binding by the Y317A ␤ IL-3 receptor was more severely impaired (ϳ20-fold lower affinity than wild-type ␤ IL-3 ), whereas the I320A ␤ IL-3 did not detectably bind mIL-3.
These results implicate Ile 320 as a key mIL-3 binding determinant and suggest that Tyr 317 plays an important additional role in mIL-3 binding.
The residues of the predicted domain 4 BЈ-CЈ loop were individually mutated to alanine, and the effect of the mutations on low affinity mIL-3 binding was assessed. Although previous work has implicated this loop in binding, to date no systematic analysis of the binding determinants in this loop has been performed. The loop comprises the residues Lys 344 , Ile 345 , Pro 346 , Lys 347 , Tyr 348 , and Ile 349 . The size and amino acid sequence of this loop differ from the BЈ-CЈ loops of the murine and human ␤c subunits (Fig. 2) suggesting potential involvement in low affinity mIL-3 binding. Pro 346 was not mutated to alanine in this study, as this substitution is likely to result in major structural perturbation of the BЈ-CЈ loop making any loss of binding ambiguous. The K344A and I345A ␤ IL-3 subunits bound mIL-3 with near wild-type affinity. However, the K347A and I349A ␤ IL-3 subunits showed a reduction in affinity for mIL-3 binding of 2-and 6-fold, respectively (Table II). The result with K347A is in agreement with the properties of a similar K347S mutant ␤ IL-3 subunit studied previously (32). Complete abolition of mIL-3 binding was found with the Y348A mutant ␤ IL-3 subunit, indicating a major role for Tyr 348 in mIL-3 binding.
Alanine substitution mutants of residues of the FЈ-GЈ loop in domain 4 of ␤ IL-3 , Ile 398 , Ser 399 , Asp 400 , Tyr 401 , and Asp 402 were prepared and examined for mIL-3 binding. Both the I398A and S399A mutant ␤ IL-3 subunits exhibited approximately wildtype low affinity mIL-3 binding. The D400A and D402A mutants bound mIL-3 with respective 4.5-and 3.5-fold reductions in mIL-3 binding affinity compared with wild-type ␤ IL-3 . The role of Asp 400 in mIL-3 binding has been examined in a previous study, as it differs from the Asn present in the m␤c receptor (32). The binding observed with D400A is consistent with this previous study, where the D400N mutant ␤ IL-3 exhibited an 8-fold reduction in mIL-3 low affinity binding (32). The Y401A mutant ␤ IL-3 did not detectably bind mIL-3, implicating Tyr 401 as a key residue for low affinity mIL-3 binding.
In order to determine whether the phenolic hydroxyl groups of the three critical Tyr residues (Tyr 21 , Tyr 348 , and Tyr 401 ) were involved in binding, phenylalanine mutants were examined (Table II). Two different effects were found. Y21F ␤ IL-3 showed impaired binding of mIL-3 (a 6-fold reduction), indicating that the hydroxyl group of Tyr 21 plays an important role in mIL-3 binding. In contrast, mIL-3 binding by the Y348F and Y401F ␤ IL-3 subunits was not severely affected. Y348F ␤ IL-3 exhibited only a small reduction in affinity for mIL-3 (less than 2-fold), whereas the Y401F ␤ IL-3 subunit bound mIL-3 with near wild-type affinity.
High Affinity mIL-3 Binding Properties of Mutant ␤ IL-3 Subunits-To determine whether the residues critical for direct mIL-3 binding were also involved in high affinity binding, COS7 cells were transiently co-transfected with expression vectors encoding the mIL-3 ␣ receptor and wild-type or mutant a Dissociation constants (K d ) Ϯ S.E. were determined from the coanalysis of data obtained from separate binding experiments by using LIGAND (37).
b Dash indicates no binding was detected above the background nonspecific binding.

FIG. 4. Scatchard representation of mIL-3 binding assays.
A depicts a representative hot saturation binding assay performed on COS7 cells co-expressing the wild-type mIL-3 ␣ and ␤ IL-3 subunits; B depicts a representative cold competition binding assay performed on COS7 cells expressing the wild-type ␤ IL-3 subunit alone. ␤ IL-3 subunits, and hot saturation binding assays were performed in which increasing amounts of 125 I-labeled mIL-3 were added (Table III).
COS7 cells transfected with the mIL-3 ␣ subunit alone exhibited one binding site of low affinity (K d , 45 nM), with a high associated error due to low absolute binding of radioiodinated mIL-3 (data not shown). Cells co-expressing the mIL-3␣ and wild-type ␤ IL-3 subunits exhibited two binding sites (Fig. 4), a low affinity site (K d ϳ45 nM) and a high affinity site (K d , 209 pM). These dissociation constants are consistent with those reported for this receptor system (14). The residues critical for mIL-3 high affinity binding have not been investigated previously.
The mutant ␤ IL-3 receptors, in which the residues in the domain 1 A-B and E-F loops, the interdomain loop, and the domain 4 BЈ-CЈ and FЈ-GЈ loops were substituted with alanine, were examined for their involvement in high affinity mIL-3 binding (Table III). Most interesting, all of these mutants were capable of binding mIL-3 with high affinity. Moreover, it is surprising that the ␤ IL-3 residues, Tyr 21 , Phe 85 , Asn 87 , Ile 320 , Tyr 348 , and Tyr 401 , which are critically required for low affinity mIL-3 binding in the absence of the mIL-3 ␣ subunit, are not obligatory for high affinity mIL-3 binding in the presence of the mIL-3 ␣ subunit. Very small reductions in high affinity binding were observed with Ala mutants of each of these residues with the exception of Tyr 348 (Table III). A slightly more pronounced reduction in high affinity mIL-3 binding (Ͼ2-fold) was observed for alanine substitution mutants of the two domain 4 BЈ-CЈ loop residues, Lys 347 and Ile 349 , two residues that play only a minor role in the direct binding of mIL-3 by ␤ IL-3 . However, none of the reductions were of sufficient magnitude to indicate critical involvement of any of the residues in the elbow region in the high affinity complex. The possibility that these residues may act synergistically to bind mIL-3 with high affinity was further probed. Studies of high affinity hIL-3 binding by h␤c have demonstrated that individually the residues in the domain 4 BЈ-CЈ loop are not obligatory for hIL-3 binding (28) but in combination play a role in high affinity hIL-3 binding (38). Likewise, studies of the human IL-4 ␣ receptor have demonstrated the additive effects of double mutations of functional epitope residues on ligand binding (25). In an analogous fashion, ␤ IL-3 constructs were prepared that contain combinations of mutations that abolish direct mIL-3 binding and cause reductions in high affinity mIL-3 binding. These ␤ IL-3 mutants, Y21A/Y401A and F85A/Y401A, were co-expressed with mIL-3 ␣ in COS7 cells, and their ability to bind mIL-3 with high affinity was assessed (Table III). Both of these double mutants were still capable of high affinity mIL-3 binding. The Y21A/ Y401A ␤ IL-3 subunit exhibited near wild-type high affinity binding, and the combination of mutations did not cause a reduction in mIL-3 binding affinity compared with the individual alanine substitutions. The F85A/Y401A ␤ IL-3 subunit exhibited a reduction in affinity for mIL-3 that is consistent with the cumulative effects of the component F85A and Y401A substitutions. Thus surprisingly, no alanine substitutions of key residues in the functional epitope for direct IL-3 binding in ␤ IL-3 were found to eliminate high affinity mIL-3 binding.
Cell-surface Expression of Mutant ␤ IL-3 Subunits-In order to verify that the reduced low and high affinity mIL-3 binding observed in assays of the K347A, I349A, and F85A/Y401A mutant ␤ IL-3 subunits was not a consequence of perturbation of the receptor translation, folding, or presence on the cell surface, flow cytometry was used to detect ␤ IL-3 expression on COS7 cells. COS7 cells, which were transiently transfected with cDNAs encoding these mutant ␤ IL-3 subunits, were stained with the anti-Aic2 monoclonal antibody (39) and a a All ␤ IL-3 mutants exhibited high affinity binding. When a large number of high affinity binding sites were present, less data were recorded for the low affinity site, and consequently a two-site model was not statistically significant. When a two-site model was found to be statistically significant (p Ͻ 0.05), the low affinity site K d was fixed as 45 nM to obtain a more accurate estimate of the K d for high affinity binding.
phycoerythrin-conjugated secondary antibody and examined by indirect flow cytometry (Fig. 5). The K347A, I349A, and F85A/Y401A mutant ␤ IL-3 subunits were found to be expressed at approximately the same level as wild-type ␤ IL-3 , with only small variations observed because of differences in transfection efficiency. The Y21A/Y401A ␤ IL-3 subunit was included for comparison and was likewise found to be expressed at a level comparable with the wild-type ␤ IL-3 subunit (Fig. 5). COS7 cells transfected with 12.5 g of the empty vector, pcEX-V3-Xba, exhibited no increase in fluorescence intensity above background observed when stained with secondary antibody alone. Additionally, reductions in low affinity mIL-3 binding by other ␤ IL-3 mutants (Table II) were not caused by perturbations to receptor translation, folding, or the presence on the cell surface, as these receptors were capable of near wild-type high affinity mIL-3 binding (Table III). Growth Response of the Y21A Mutant ␤ IL-3 Subunit to mIL-3-The finding that all of the ␤ IL-3 alanine substitution mutants examined in the present work were able to bind mIL-3 with high affinity was surprising, as alanine substitutions of the analogous residues in the h␤c receptor abolished high affinity ligand binding. In order to further probe this finding, we examined the ability of the Y21A mutant ␤ IL-3 subunit to be activated and to deliver a downstream signal in response to mIL-3 by using cell proliferation as an absolute measure. This mutant ␤ IL-3 was chosen for this study because of its obligatory involvement in direct mIL-3 binding and non-involvement in high affinity mIL-3 binding. Wild-type or mutant mIL-3 receptors were stably transfected into the mIL-2-dependent lymphoid cell line, CTLL-2, to examine its capacity to grow in response to mIL-3 in two steps. Initially a cDNA encoding the mIL-3 ␣ subunit was stably transfected into CTLL-2 cells via the pEF-IRES-N vector, which encodes G418 antibiotic resistance. The wild-type or Y21A m␤ IL-3 subunits were subsequently transfected into G418-resistant cells by using the vector, pEF-IRES-P, which encodes puromycin resistance. Cells resistant to these antibiotics, which had been maintained in mIL-2, were then used in proliferation assays to determine their responsiveness to mIL-3 (Fig. 6A). As evident from the curve in Fig. 6A, the mIL-3-stimulated growth responses of the CTLL-2 cell lines co-expressing mIL-3 ␣ and wild-type ␤ IL-3 or Y21A ␤ IL-3 are comparable, whereas the parent cell line, CTLL-2 mIL-3␣, did not detectably respond to mIL-3. This verifies that the high affinity binding observed for Y21A ␤ IL-3 results in normal receptor activation.
Residues of h␤c Forming the Functional Epitope for High Affinity hIL-3 Binding-Previous studies (31) in this laboratory demonstrated that alanine mutations of the domain 1 residues, Tyr 15 and Phe 79 , of the h␤c receptor abolished hGM-CSF binding and severely impaired receptor activation by hGM-CSF and hIL-5. In view of the surprising results with high affinity binding of the ␤ IL-3 receptor, the role of these residues in high affinity hIL-3 binding and receptor activation by hIL-3 was examined.
COS7 cells were co-transfected with expression vectors encoding wild-type, Y15A, or F79A h␤c and wild-type hIL-3 ␣ subunits. Saturation binding assays were performed on these transiently transfected cells by using increasing amounts of 125 I-radiolabeled hIL-3, and the dissociation constants were determined (Table IV). COS7 cells expressing the wild-type h␤c and hIL-3␣ receptors exhibited two binding sites. In agreement with earlier studies, it was found that accurate estimation of the low affinity site is not possible because of its very low K d values (ϳ100 nM) resulting in a high associated error (28). Accordingly, the low affinity site K d was fixed at 100 nM during analysis in LIGAND in order to obtain a more accurate estimate of the high affinity site (K d , 107 pM). The dissociation constants we obtained are consistent with those reported in previous studies (28,30) on the hIL-3 receptor. The h␤c subunit used in these studies was derived from HL-60 eosinophils and contained a six-amino acid insertion in the C-D loop of domain 3 (34). We have shown that this does not affect wild-type high affinity binding of GM-CSF (31), and here we demonstrated that there was no effect on normal hIL-3 high affinity binding.
Alanine substitution of Tyr 15 (A-B loop) and Phe 79 (E-F loop) in domain 1 of h␤c abolished high affinity binding (Table  IV), implicating both of these residues in the formation of the high affinity hIL-3 complex. Because the formation of the high affinity hIL-3 complex is believed to be necessary for receptor activation and downstream signaling, we reconstituted the wild-type and mutant hIL-3 receptors in the mIL-2-dependent lymphoid cell line, CTLL-2, to correlate the loss of high affinity binding with an unequivocal signaling response, proliferation. Initially, CTLL-2 cells were transfected with a cDNA encoding the hIL-3 ␣ subunit, and subsequently the wild-type or mutant ␤ IL-3 subunit was introduced via vectors encoding antibiotic resistance (as described for the mIL-3 receptors above). Cells resistant to these antibiotics, which had been maintained in mIL-2, were then used in proliferation assays to determine their responsiveness to hIL-3 (Fig. 6B).
The amount of hIL-3 resulting in half-maximal stimulation of the cell line co-expressing wild-type h␤c and hIL-3 ␣ was set at 1 unit, and the relative amounts of hIL-3 required for 50% stimulation of cell lines expressing mutant receptors was estimated (Table IV and Fig. 6B). Y15A and F79A mutant h␤c receptors were at least 64-and 5-fold, respectively, less responsive to hIL-3 confirming that the abolition of high affinity binding was reflected in reduced efficiency of signaling. As is evident in Fig. 6B, the proliferation curves of CTLL-2 cells containing the hIL-3 ␣ and Y15A and F79A h␤c subunits did not reach a plateau in this hIL-3 titration because of limitations on the volume of hIL-3 that could be applied. Conse- quently, the reductions in hIL-3 responsiveness observed for cell lines expressing the mutant h␤c receptors are likely to be underestimated. The parent cell line, CTLL-2 hIL-3␣, did not respond to hIL-3 even at doses in excess of 200 units (Fig. 6B). The above results together with previous mutagenesis studies on domain 4 (30,38) suggest that the functional epitope of the h␤c subunit for high affinity hIL-3 binding consists of Tyr 15 and Phe 79 in domain 1 and Tyr 403 in domain 4, with secondary contributions from Tyr 347 , His 349 , and Ile 350 . The reductions in high affinity binding and receptor activation by hIL-3 observed for the Y15A and F79A mutant h␤c subunits are not a consequence of impaired receptor translation, folding, or cell-surface expression, as previously these mutants were shown to be expressed at wild-type levels in COS7 cells (31). DISCUSSION A conserved ligand-binding interface has been identified for several members of the hematopoietin receptor superfamily including the growth hormone (18,21,22), erythropoietin (19,23,24), and IL-4 ␣ receptors (20,25). These ligand-binding interfaces are present at an elbow of ϳ90°that is formed between the two fibronectin III domains constituting the extracellular domains of these receptors. Residues are contributed from a combination of the A-B and E-F loops of the membrane-distal domain and the BЈ-CЈ and FЈ-GЈ loops of the membrane-proximal domain to form a cluster of residues critical for ligand binding. The crystal structure of the h␤c receptor showed that a similar elbow, representing a potential ligandbinding interface, is formed between domains 1 and 4 of the two different monomers that compose the h␤c homodimer (26). This elbow region resembled the ligand-binding interface conserved in the simpler two-domain hematopoietin receptors, despite domains 1 and 4 being non-contiguous. Indeed, mutagenesis studies suggest that residues in domains 1 and 4 of the h␤c subunit form the functional epitope for human GM-CSF and IL-5 binding in a manner analogous to the simpler hematopoietin receptors (31, 28 -30). In the present work we show that this h␤c functional epitope is similarly involved in hIL-3 high affinity binding and receptor activation. The residues, Tyr 15 and Phe 79 , of domain 1 in h␤c were shown to be critical for high affinity hIL-3 binding and also required for optimal receptor activation. As discussed in detail previously (31), it is possible that the effect of the F79A mutation of h␤c could be due to an indirect effect on Tyr 15 . A detailed understanding of the mechanism of binding awaits the determination of the structure of Plots depict the counts per min of tritiated thymidine incorporated into the DNA of CTLL-2 cells stably expressing wild-type or mutant receptors in response to a serial dilution of growth factor. A shows typical data for CTLL-2 cells expressing mIL-3 ␣ alone (circle) and with wild-type (diamond) or Y21A (triangle) ␤ IL-3 subunits. B shows typical data for the hIL-3 stimulation of CTLL-2 cells expressing hIL-3 ␣ alone (circle) and with wild-type (diamond), Y15A (triangle), or F79A (square) h␤c. a Proliferation assays were performed on CTLL-2 cells co-expressing the hIL-3 ␣ and indicated ␤c subunits. Data for these proliferation assays are shown in Fig. 6B. b A low affinity site of 100 nM was fitted in LIGAND (37) to enable a more accurate estimation of the high affinity binding site.
c No high affinity binding sites were detected.
the high affinity hIL-3 complex, but the functional epitope based on this and previous mutagenesis studies (30,38) consists of Tyr 15 , Phe 79 , and Tyr 403 with secondary interactions from residues Tyr 347 to Ile 350 in the BЈ-CЈ loop. The residues in the murine ␤ IL-3 receptor critical for mIL-3 binding were also determined. This receptor provides the opportunity to compare residues involved in both low and high affinity binding. The murine ␤ IL-3 and h␤c receptors are highly homologous in sequence suggesting that they may have homologous structures. As part of the present work, cross-linking and ultracentrifugation studies provided evidence that the expressed extracellular domains of the ␤ IL-3 and m␤c receptors, like that of the h␤c receptor, exist as homodimers in the absence of ligand. Taken together these findings suggest the ␤ IL-3 and m␤c receptors may be intertwined homodimers like h␤c, predicting an involvement of domains 1 and 4 in forming the ligand-binding epitope.
Mutagenesis of the ␤ IL-3 receptor revealed that Ala substitutions of Tyr 21 (A-B loop), Phe 85 , and Asn 87 (E-F loop) of domain 1; Ile 320 of the interdomain loop; and Tyr 348 (BЈ-CЈ loop) and Tyr 401 (FЈ-GЈ loop) of domain 4 resulted in no detectable direct low affinity mIL-3 binding. We estimate that the dynamic range of the assay used extends to slightly beyond ϳ370 nM, or a little over 20 times the K d value of wild-type ␤ IL-3 for direct binding. The alanine substitutions of several other ␤ IL-3 residues also impaired low affinity mIL-3 binding. Arg 84 and Tyr 317 were shown to have major secondary roles. Asp 400 and Asp 402 , which neighbor the critical residue Tyr 401 , also had detectable effects. The critical ␤ IL-3 residues, Tyr 21 , Phe 85 , Tyr 348 , and Tyr 401 , align with 4 of the 6 residues of h␤c shown to be involved in high affinity ligand binding (Fig. 2). Thus direct binding involves the cooperation of domains 1 and 4 in forming an analogous ligand-binding interface in ␤ IL-3 to that shown for h␤c (Fig. 1).
To understand further the role of Tyr 21 , Tyr 348 , and Tyr 401 of ␤ IL-3 in direct mIL-3 binding, phenylalanine mutants of these residues were prepared. Y21F ␤ IL-3 was able to bind mIL-3 with 6-fold lower affinity than wild-type ␤ IL-3 , indicating that Phe was able to restore some mIL-3 binding but that the phenolic hydroxyl group plays an important role. In contrast, Y348F ␤ IL-3 was able to bind mIL-3 with an affinity within 2-fold of the wild-type receptor, whereas Y401F ␤ IL-3 exhibited near wildtype affinity indicating that the phenolic hydroxyl is not critical in either case. In comparison, with hGM-CSF high affinity binding to h␤c, phenylalanine substitution mutants of the critical tyrosines Tyr 15 and Tyr 347 were fully active (29,31), whereas high affinity hGM-CSF binding was abolished by the Y401F mutation. 2 An unexpected finding in the present work is the major differences in the interactions between mIL-3 and ␤ IL-3 in low (direct) and high affinity mIL-3 binding. None of the residues that are individually critical for direct IL-3 binding are obligatory for high affinity binding. The simplest prediction would have been that the residues critical for high affinity binding would be those involved in low affinity binding or that the latter residues might be involved in an obligatory intermediate step in the formation of the high affinity complex with the mIL-3␣ subunit. It would therefore appear that the mechanism of high affinity IL-3 binding in the murine ␤ IL-3 receptor is fundamentally different from that of its close relative the h␤c receptor. The possibility of an additional signaling role for the direct IL-3 binding site of the murine ␤ IL-3 receptor remains unresolved at present.